A pattern global edge intensity quantification method and device for garment patterns

Abstract

The application discloses a pattern global edge intensity quantification method and device for garment patterns, and comprises the following steps: down-sampling a pattern of a garment pattern to obtain patterns of different scales; performing edge intensity calculation on each scale of the pattern based on a Laplacian operator to obtain an edge intensity map, and performing quantile statistics on the edge intensity map to obtain edge intensity description features; and splicing the edge intensity description features corresponding to all scales to obtain global edge intensity description features. The application aims to overcome the limitations of existing methods in style distinguishing and clarity judging of patterns, thereby providing a more accurate and statistical edge feature for the image processing field.

Description

Technical Field

[0001] This invention belongs to the field of pattern processing technology, specifically relating to a method and apparatus for global edge intensity quantization of clothing patterns. Background Technology

[0002] Global edge intensity quantization of clothing patterns plays a crucial role in clothing pattern classification and recognition. Traditional edge detection operators have significant shortcomings in assessing the sharpness of image edges. They often only focus on the presence of edges, ignoring the global sharpness of the edges. When processing images with different styles, such as line drawing and ink painting, existing operators cannot provide a global assessment of the image style.

[0003] The Laplacian operator is a two-dimensional isotropic measurement of the second spatial derivative of an image. The Laplacian of an image highlights regions of rapidly changing intensity and is therefore commonly used for edge detection (see Zero-Crossing Edge Detector). This Laplacian operator typically takes a single-channel grayscale image as input and generates another grayscale image as output, but it is not suitable for evaluating the global sharpness of edges.

[0004] Existing edge detection operators such as Sobel and Canny detect edges by calculating the gradient of an image. However, these edge detection operators typically only detect the presence of edges, while struggling to assess the global sharpness of edges.

[0005] Patent application CN110321762A discloses a high-precision recognition method for contour recognition, and patent application CN110335226A discloses an image recognition method for laboratory environments. Both of these technical solutions use multi-scale decomposition and spatial domain gradient operators for edge detection. The extracted edge features are used for contour recognition and image recognition. However, like most existing technologies, these two technical solutions can only detect the existence of edges without analyzing the edge strength and clarity.

[0006] When faced with complex patterns and diverse clothing designs, existing edge detection operators often struggle to provide comprehensive pattern edge information, particularly failing to offer global quantification for overall image quality and content realism. Furthermore, existing algorithms have significant limitations in distinguishing between images of different styles and assessing the sharpness of blurred images. Therefore, the image processing field requires a more comprehensive and accurate method for pattern edge intensity quantification to meet the needs of global evaluation and sharpness assessment for images of different styles. Summary of the Invention

[0007] In view of the above, the purpose of this invention is to provide a method and apparatus for global edge intensity quantification of clothing patterns, aiming to overcome the limitations of existing methods in distinguishing styles and judging sharpness of patterns, thereby providing a more accurate and statistically significant edge feature for the field of image processing.

[0008] To achieve the above-mentioned objectives, this invention provides a method for global edge intensity quantization of clothing patterns, comprising the following steps:

[0009] Downsampling of clothing patterns yields patterns at different scales;

[0010] Edge intensity maps are obtained by calculating the edge intensity of patterns at each scale based on the Laplacian operator, and edge intensity descriptive features are obtained by performing quantile statistics on the edge intensity maps.

[0011] The global edge intensity description features are obtained by concatenating the edge intensity description features corresponding to all scales.

[0012] Preferably, an edge intensity map is obtained by calculating the edge intensity of the pattern at each scale based on the Laplacian operator, including:

[0013] The colored pattern is converted into a grayscale image, and the edge intensity is calculated on the grayscale image using the Laplace operator to obtain an edge intensity map with the same size as the pattern. In the edge intensity map, each pixel value is the second-order gradient value of the Laplace operator with respect to the corresponding pixel position in the grayscale image.

[0014] Preferably, the edge intensity descriptive features are obtained by performing statistical quantile analysis on the edge intensity map, including:

[0015] Take K quantiles of a specified proportion, and statistically analyze all pixel values ​​in the edge intensity map based on the quantiles to obtain K pixel values ​​that satisfy the K quantile division to form the edge intensity description feature.

[0016] Preferably, the K pixel values ​​that satisfy the K quantile division mean that all pixel values ​​in the edge intensity map are considered as a pixel distribution, and the pixel distribution is divided into two parts that satisfy the proportion corresponding to a single quantile according to the division point. The pixel value corresponding to this division point is the single pixel value that satisfies the single quantile division. The corresponding K pixel values ​​are obtained by dividing the pixel distribution K times with K quantiles.

[0017] Preferably, different numbers of quantiles are set for patterns at each scale, and the set quantile values ​​vary.

[0018] To achieve the above-mentioned objectives, this invention also provides a global edge intensity quantization device for clothing patterns, including a scale division module, a feature extraction module, and a feature integration module.

[0019] The scale division module is used to downsample the pattern of the clothing design to obtain patterns of different scales;

[0020] The feature extraction module is used to calculate the edge intensity of the pattern at each scale based on the Laplacian operator to obtain an edge intensity map, and to perform quantile statistics on the edge intensity map to obtain edge intensity description features;

[0021] The feature integration module is used to concatenate the edge intensity description features corresponding to all scales to obtain the global edge intensity description features.

[0022] Preferably, in the feature extraction module, edge intensity maps are obtained by calculating the edge intensity of the pattern at each scale based on the Laplacian operator, including:

[0023] The colored pattern is converted into a grayscale image, and the edge intensity is calculated on the grayscale image using the Laplace operator to obtain an edge intensity map with the same size as the pattern. In the edge intensity map, each pixel value is the second-order gradient value of the Laplace operator with respect to the corresponding pixel position in the grayscale image.

[0024] Preferably, in the feature extraction module, obtaining edge intensity description features by performing statistical quantile analysis on the edge intensity map includes:

[0025] Take K quantiles of a specified proportion, and statistically analyze all pixel values ​​in the edge intensity map based on the quantiles to obtain K pixel values ​​that satisfy the K quantile division to form the edge intensity description feature.

[0026] To achieve the above-mentioned objectives, embodiments of the present invention also provide a computing device, including a memory and one or more processors, wherein the memory stores executable code, and the one or more processors execute the executable code to implement the above-mentioned pattern global edge intensity quantization method for clothing patterns.

[0027] To achieve the above-mentioned objectives, embodiments of the present invention also provide a computer-readable storage medium storing a program thereon, which, when executed by a processor, implements the above-mentioned method for global edge intensity quantization of clothing patterns.

[0028] Compared with the prior art, the beneficial effects of the present invention include at least the following:

[0029] Compared to traditional edge detection algorithms (such as Sobel, Canny, etc.), which tend to extract features from local regions, the method of this invention achieves a more comprehensive description of clothing patterns through multi-scale global edge intensity quantization, which can better capture the overall structure of clothing patterns and provide richer information for subsequent analysis.

[0030] The method of this invention can adapt to different design ideas for patterns, thus performing better when dealing with garment patterns of various sizes and dimensions. This is crucial for handling subtle textures and elements of different granularities in garment patterns, improving the method's performance when dealing with patterns with detail and diversity, and is applicable to the evaluation of garment patterns in different locations (such as collar, hem, and back).

[0031] The introduction of adjustable parameters such as the size of the Laplacian operator, the numerical value of quantiles, the number of features, and the dimension of features, along with consideration of multi-scale analysis, provides the method with the possibility of optimization in specific garment pattern domains. By adjusting the parameters, it can better adapt to specific garment patterns and styles, improving the practical applicability of the method and enhancing its flexibility and versatility in different scenarios.

[0032] By combining pixel values ​​corresponding to multiple quantiles into edge intensity description features, a more informative global edge description is provided. Clothing patterns typically possess a specific style and structure as a whole. Compared to traditional algorithms that use simpler feature representations, this complex edge intensity description feature is better able to capture the intricate structure of clothing patterns at different scales.

[0033] The method utilizes statistical quantiles, demonstrating robustness against noise in images and exhibiting greater robustness compared to some traditional noise-sensitive algorithms. Furthermore, it still provides reliable edge descriptions even when faced with flower-shaped deformations. Attached Figure Description

[0034] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0035] Figure 1 This is a flowchart of a method for global edge intensity quantization of clothing patterns provided in an embodiment of the present invention;

[0036] Figure 2 These are clothing patterns provided in the embodiments of the present invention, wherein (a) is a geometric pattern, (b) is a rendered pattern, and (c) is a snow mountain pattern;

[0037] Figure 3 This is a scatter plot provided in an embodiment of the present invention;

[0038] Figure 4 This is a schematic diagram of the structure of the global edge intensity quantization device for clothing patterns provided in an embodiment of the present invention;

[0039] Figure 5 This is a schematic diagram of the structure of the computing device provided in an embodiment of the present invention. Detailed Implementation

[0040] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and do not limit the scope of protection of this invention.

[0041] The technical concept of this invention is as follows: Existing edge detection methods only focus on the existence of edges, but ignore the global sharpness of the edges. To solve this technical problem, this invention provides a global edge intensity quantization scheme for clothing patterns. By using a multi-scale analysis method based on the Laplacian operator, edge gradients are calculated hierarchically and at different scales, and quantile statistics of edge intensity are performed. This aims to more accurately capture the edge intensity information of patterns at different levels, which can be applied to distinguish image styles (such as simple line drawing style and ink painting style) and to determine the sharpness of images.

[0042] like Figure 1 As shown in the embodiment, a method for global edge intensity quantization of clothing patterns is provided, which includes the following steps:

[0043] S110, downsampling the pattern of the clothing design to obtain patterns of different scales.

[0044] In this embodiment, patterns of different scales are obtained by downsampling the clothing pattern, forming an image pyramid. The number of layers in the pyramid is set to L, and the image size of the i-th layer is M. i ×N i The specific downsampling process can be expressed by the following formula:

[0045]

[0046] Where M and N are the dimensions of the original pattern, and the number of pyramid layers L can be selected according to application requirements.

[0047] S120: For each scale of the pattern, the edge intensity is calculated based on the Laplacian operator to obtain the edge intensity map, and the edge intensity map is statistically analyzed by quantiles to obtain the edge intensity description features.

[0048] To address the lack of quantization calculations for global edge strength in traditional feature detection operators, Laplacian edge detection and quantile statistics are employed to obtain edge strength descriptive features. The specific inputs are the Laplacian operator size (kernel_size), the feature vector dimension (K), and the K quantile ratios.

[0049] Then, the colored pattern is converted into a grayscale image, and the edge intensity is calculated on the grayscale image using the Laplace operator to obtain an edge intensity map with the same size as the pattern. In the edge intensity map, each pixel value is the second gradient value of the Laplace operator with respect to the corresponding pixel position in the grayscale image. The larger the second gradient value, the more obvious the edge.

[0050] Finally, K quantiles of a specified proportion are selected, and all pixel values ​​in the edge intensity map are statistically analyzed based on these quantiles to obtain K pixel values ​​that satisfy the K quantile division, forming the edge intensity description feature. Here, the K pixel values ​​satisfying the K quantile division mean that all pixel values ​​in the edge intensity map are considered as a pixel distribution, and this distribution is divided into two parts according to a dividing point that satisfies the proportion corresponding to a single quantile. The pixel value corresponding to this dividing point is the single pixel value satisfying the single quantile division. K divisions of the pixel distribution using K quantiles yield the corresponding K pixel values.

[0051] It should be noted that different numbers of quantiles can be set for each scale of the pattern, and the set quantile values ​​can also vary. The set quantiles are used to find the gradient threshold to distinguish the pixel distribution corresponding to the edge intensity map into strong edges and weak edges, which can more clearly capture the edge intensity information of patterns of different sizes.

[0052] S130: The edge intensity description features corresponding to all scales are concatenated to obtain the global edge intensity description features.

[0053] After obtaining the edge intensity description features corresponding to all scales, they are concatenated to obtain the global edge intensity description features.

[0054] The embodiments also provide specific experiments to verify that the features obtained by edge strength calculation and quantile statistics of the Laplacian operator provided by the present invention have significantly better classification accuracy than edge features obtained by traditional methods in the classification task for edge sharpness.

[0055] Experimental data are as follows Figure 2 The three categories of clothing patterns shown exhibit significant differences in edge sharpness: 50 geometric patterns, 51 rendered patterns, and 43 snow mountain patterns. Each category was divided into a training set and a test set at an equal ratio of (0.8, 0.2).

[0056] The extracted features are divided into three categories. The first category is the global edge intensity description feature edge_vectors extracted by this invention, which has 10 dimensions and uses quantiles [0.5, 0.6, 0.7, 0.8, 0.85, 0.9, 0.92, 0.95, 0.97, maximum value]. The second category is the calculated gray-level co-occurrence matrix glcm_vectors, which has 72 dimensions. The gray-level co-occurrence matrix glcm_vectors takes 3 distances (1, 2, 4), 4 angles (0, pi / 4, pi / 2, 3*pi / 4), and 6 indices ('contrast', 'dissimilarity', 'homogeneity', 'energy', 'correlation', 'ASM'), thus totaling 72 dimensions (72 = 3*4*6). The third category is the comprehensive feature comb_feature, which is obtained by concatenating edge_vectors and glcm_vectors, totaling 82 dimensions.

[0057] The classification model used for the classification task is the SVC model, where the kernel is the RBF kernel, with parameters C=1 and gamma="scale". The corresponding classification results are as follows:

[0058] Using edge vectors as features for classification with the SVC model:

[0059] Training set classification accuracy: 0.6783;

[0060] Test set classification accuracy: 0.6552;

[0061] Using glcm_vectors as features for classification with the SVC model:

[0062] Training set classification accuracy: 0.4522;

[0063] Test set classification accuracy: 0.3103;

[0064] Using comb_feature as a feature for classification with the SVC model:

[0065] Training set classification accuracy: 0.6261;

[0066] The classification accuracy of the test set is 0.4138.

[0067] Based on the above classification results, it can be concluded that, compared with traditional methods, the edge intensity description features extracted by the method of this invention can effectively improve the classification accuracy for image edge sharpness. Furthermore, from... Figure 3 The scatter plot shown demonstrates that the method of this invention can significantly distinguish between multiple types of images with visible differences in edge sharpness, achieving the desired effect.

[0068] like Figure 4 As shown in the embodiment, a global edge intensity quantization device 500 for clothing patterns is also provided, including a scale division module 510, a feature extraction module 520, and a feature integration module 530. The scale division module 510 is used to downsample the clothing pattern to obtain patterns of different scales. The feature extraction module 520 is used to calculate the edge intensity of each scale pattern based on the Laplacian operator to obtain an edge intensity map, and to perform quantile statistics on the edge intensity map to obtain edge intensity description features. The feature integration module 530 is used to stitch together the edge intensity description features corresponding to all scales to obtain global edge intensity description features.

[0069] In the feature extraction module 520, edge intensity maps are obtained by performing edge intensity calculation based on the Laplacian operator for each scale of the pattern. This includes converting the colored pattern into a grayscale image and calculating the edge intensity on the grayscale image using the Laplacian operator to obtain an edge intensity map with the same size as the pattern. In the edge intensity map, each pixel value is the second-order gradient value of the Laplacian operator with respect to the corresponding pixel position in the grayscale image.

[0070] In the feature extraction module 530, the edge intensity map is statistically divided into quantiles to obtain edge intensity description features, including: taking K quantiles of a specified proportion, and statistically analyzing all pixel values ​​in the edge intensity map based on the quantiles to obtain K pixel values ​​that satisfy the K quantile division to form edge intensity description features.

[0071] It should be noted that the global edge intensity quantization device for clothing patterns provided in the above embodiments should be illustrated using the above-described functional module division as an example when performing global edge intensity quantization for clothing patterns. The functions can be assigned to different functional modules as needed, i.e., the internal structure of the terminal or server can be divided into different functional modules to complete all or part of the functions described above. Furthermore, the global edge intensity quantization device for clothing patterns provided in the above embodiments and the global edge intensity quantization method for clothing patterns belong to the same concept. The specific implementation process is detailed in the global edge intensity quantization method for clothing patterns, and will not be repeated here.

[0072] The embodiment also provides a computing device, including a memory and one or more processors, wherein the memory stores executable code, and when the one or more processors execute the executable code, it is used to implement the above-described pattern global edge intensity quantization method for clothing patterns, including the following steps:

[0073] S110, downsampling the pattern of the clothing design to obtain patterns of different scales;

[0074] S120, perform edge intensity calculation based on the Laplacian operator for the pattern at each scale to obtain an edge intensity map, and perform quantile statistics on the edge intensity map to obtain edge intensity description features;

[0075] S130: The edge intensity description features corresponding to all scales are concatenated to obtain the global edge intensity description features.

[0076] like Figure 5 As shown, the computing device provided in this embodiment, at the hardware level, includes not only a processor and memory, but also an internal bus, network interface, memory, and other hardware required for business operations. The memory is non-volatile memory. The processor reads the corresponding computer program from the non-volatile memory into memory and then runs it to implement the pattern global edge intensity quantization method for clothing patterns described in S110-S130 above. Of course, in addition to the software implementation, this invention does not exclude other implementation methods, such as logic devices or a combination of hardware and software, etc. That is to say, the execution subject of the following processing flow is not limited to each logic unit, but can also be hardware or logic devices.

[0077] The embodiment also provides a computer-readable storage medium having a program stored thereon. When executed by a processor, the program implements the above-described pattern global edge intensity quantization method for clothing patterns, including the following steps:

[0078] S110, downsampling the pattern of the clothing design to obtain patterns of different scales;

[0079] S120, perform edge intensity calculation based on the Laplacian operator for the pattern at each scale to obtain an edge intensity map, and perform quantile statistics on the edge intensity map to obtain edge intensity description features;

[0080] S130: The edge intensity description features corresponding to all scales are concatenated to obtain the global edge intensity description features.

[0081] In this embodiment, computer-readable media includes both permanent and non-permanent, removable and non-removable media, and information storage can be implemented by any method or technology. Information can be computer-readable instructions, data structures, program modules, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic magnetic disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include transient computer-readable media, such as modulated data signals and carrier waves.

[0082] The specific embodiments described above illustrate the technical solution and beneficial effects of the present invention in detail. It should be understood that the above description is only the most preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, additions, and equivalent substitutions made within the scope of the principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A pattern global edge intensity quantification method for garment pattern, characterized in that, Includes the following steps: Downsampling of clothing patterns yields patterns at different scales; Edge intensity maps are obtained by performing edge intensity calculation based on the Laplacian operator for patterns at each scale, including: converting the colored pattern into a grayscale image, and calculating the edge intensity on the grayscale image using the Laplacian operator to obtain an edge intensity map with the same size as the pattern, wherein each pixel value in the edge intensity map is the second-order gradient value of the Laplacian operator with respect to the corresponding pixel position in the grayscale image; The edge intensity map is then subjected to quantile statistics to obtain edge intensity description features, including: taking K quantiles of a specified proportion, and statistically analyzing all pixel values ​​in the edge intensity map based on the quantiles to obtain K pixel values ​​that satisfy the K quantile division to form edge intensity description features; Here, the K pixel values ​​that satisfy the K quantile division mean that all pixel values ​​in the edge intensity map are regarded as a pixel distribution, and the pixel distribution is divided into two parts that satisfy the proportion corresponding to a single quantile according to the division point. The pixel value corresponding to this division point is the single pixel value that satisfies the single quantile division. The corresponding K pixel values ​​are obtained by dividing the pixel distribution K times with K quantiles. The global edge intensity description features are obtained by concatenating the edge intensity description features corresponding to all scales.

2. The method of quantifying the global edge intensity of a pattern of a garment pattern according to claim 1, wherein, Different numbers of quantiles are assigned to patterns at each scale, and the assigned quantile values ​​vary.

3. A pattern global edge intensity quantification device for a garment pattern, characterized in that, It includes a scale division module, a feature extraction module, and a feature integration module; The scale division module is used to downsample the pattern of the clothing design to obtain patterns of different scales; The feature extraction module is used to calculate the edge intensity of the pattern at each scale based on the Laplacian operator to obtain an edge intensity map, including: converting the colored pattern into a grayscale image, and using the Laplacian operator to calculate the edge intensity on the grayscale image to obtain an edge intensity map with the same size as the pattern, wherein each pixel value in the edge intensity map is the second gradient value of the Laplacian operator with respect to the corresponding pixel position in the grayscale image; The edge intensity map is then subjected to quantile statistics to obtain edge intensity description features, including: taking K quantiles of a specified proportion, and statistically analyzing all pixel values ​​in the edge intensity map based on the quantiles to obtain K pixel values ​​that satisfy the K quantile division to form edge intensity description features; Here, the K pixel values ​​that satisfy the K quantile division mean that all pixel values ​​in the edge intensity map are regarded as a pixel distribution, and the pixel distribution is divided into two parts that satisfy the proportion corresponding to a single quantile according to the division point. The pixel value corresponding to this division point is the single pixel value that satisfies the single quantile division. The corresponding K pixel values ​​are obtained by dividing the pixel distribution K times with K quantiles. The feature integration module is used to concatenate the edge intensity description features corresponding to all scales to obtain the global edge intensity description features.

4. A computing device comprising a memory and one or more processors, the memory having stored therein executable code, wherein, When the one or more processors execute the executable code, they are used to implement the pattern global edge intensity quantization method for clothing patterns as described in claim 1 or 2.

5. A computer readable storage medium, characterized in that, A computer program product, comprising a computer readable storage medium having stored thereon instructions that, when executed by a computer, implement the garment pattern oriented global edge intensity quantification method according to claim 1 or 2.

Images